Human input vital for adapting quant models to changing markets

Major players in the systematic investment world are attempting to raise awareness of the blind reliance on “black box”, data-led models, which has led to previous fund collapses and other serious problems for quantitative investors.

The idea behind such strategies is that investment decisions are based on disciplined statistical treatment of various sources of information ranging from fundamental to behavioural and market data, thoroughly back-tested using historical data to measure performance, and executed systematically. This means taking away the subjectivity of discretionary managers.

But quant models need to be based on sound financial ideas to work well, and quality of data needs to be checked and assessed regularly, making the judgemental input of the quant fund manager a key factor.

“Quantitative techniques are a useful toolbox to test statements that people make about financial markets, but they usually hardly reveal any hidden reality,” says Nicolas Gaussel, CIO of the quantitative management team at Lyxor Asset Management.

“Letting the data talk,” as some claim, is dangerous. There may for instance be a correlation between variables such as the temperature in winter and stock price, without any real cause-effect relationship between them.

The value of this investment style lies much more in the team itself than the programme, believes Mr Gaussel, and it is paramount the model adapts to the changing market environment.

Quant managers can be naïve and fall in love with their models, believes Oscar Vermeulen, director at multi-management firm Altis Investment Management. This attitude played its part in the quant crash of 2007, where quant managers, deceived by the homogeneous and consistent period started in 2003, were betting on the same underlying momentum factors and did not understand the demand for market liquidity was closely correlated to the positive momentum in the market.

When some big players cut their positions, there was a knock-out effect on all the others. “Many big quant teams, made of tens of PhDs, had popped up at the most prestigious firms in a matter of a few years, but they had “never lived the markets”, he says.

However, he believes today’s market presents a good environment for investing in quant strategies, as a quant manager’s track record is made and tested under the “Goldilocks” period of 2003-2006, the credit crunch and crisis of 2007-8 and the recovery after dislocation of 2009-10, which gives a fund selector more scope to see how a model performs in different markets scenarios.

Also, today’s markets are much less “efficient” than before, with real dislocation in credit, and a huge dispersion in valuation in investment grade credit, which should prove a more rewarding hunting ground for quant modellers than the very homogenous, low spread market of 2006. Moreover, the incredible growth in passive and semi-passive investing over the past 15 years, up from almost nothing to 50 per cent of the cake, means index constituents are valued at least partly on momentum, not just fundamentals.

Oscar Vermeulen, Altis Investment Management

Major Role to play

“Market Cap weighting is a style with incredible momentum, which makes markets somewhat inefficient, in a manner that can be exploited by systemic valuation driven quant models,” says Mr Vermeulen. “If you are able to identify good quant managers, then quant managers have a fantastic role to play in your portfolio, because their alpha will be highly uncorrelated to market climate and therefore to the most style-exposed alphas of non quant managers, which is for us one of the most desirable characteristics,” he says.

The problem with quant funds is that they are very hard to analyse. “With discretionary managers, you can go and see the manager, the teams, their passion and commitment, but you cannot interview a quant model, you have to rely much more on the quantitative verification of alpha.”

Even if the quant manager can explain the logic of his strategy, what generates alpha is the way the algorithm is implemented. He is not going to explain the software that defines the algorithm, as it is core intellectual property, and even if he did give out the nitty gritty details, this is too hard to understand, unless you yourself are a quant manager. Manager selection teams can only look at the big picture and the output of the “black box”, by using sophisticated statistical analysis.

Within alpha quantitative strategies, it is easier to evaluate the bottom-up variations, which generate alpha from the selection of securities. “You have hundreds of stock bets every year, and so purely from a statistical point of view, you have a large sample size, and you evaluate how successful they are with their stockpicking,” says Mr Vermeulen.

Top-down alpha quant strategies, driven by macro calls, may make only a few decisions every year, and with such a small sample size, they are harder to verify. Because of these difficulties in evaluating top-down quants, the focus is on the bottom-up quant strategies, generally in markets reasonably inefficient or dislocated, and where there is enough quantitative data available. A good example today is European investment grade corporate bonds, he explains.

A bottom-up approach is less risky and for this reason is preferred by a firm like Robeco. “We want to benefit from the power of diversification. If you make a macro call, that is basically an all-or-nothing call, like timing the market, which is extremely difficult,” says David Blitz, head of quantitative equity research at the Dutch firm.

However, in a bottom-up process, there is also risk that big macro calls rise and it is important to avoid this risk by distinguishing between the macro momentum of the stock – which comes from the macro economic environment – from the stock specific momentum part.

“If you buy stocks that have high stock specific momentum, you become less sensitive to these macro movements, and as a result you reduce the risk a lot, but not the return,” says Mr Blitz.

There is some widespread scepticism about the possibility to run effective quant models in emerging or frontier markets, with the poor quality of the data and difficulty of taking into account political risks being the main concerns. However, Mr Blitz believes these fears are largely unjustified.

“We run quant strategies both in developed and emerging markets, and emerging markets quant models are more effective – the lower data quality and the larger country specific risk don’t seem to affect their effectiveness.”

This may be also a consequence of the lack in popularity of this strategy, which leaves more inefficiencies in the market to be exploited. More recently, the firm successfully tested quant strategies in frontier markets, like Vietnam, Ukraine and African countries. “We spent more than eleven months in gathering good data but just a couple of weeks to test the strategy,” adds Mr Blitz.

Even acknowledging the diversification benefit of employing quant funds in portfolios, major private banks find little client interest in these strategies.

“We find the audience for quantitative strategies in the private wealth client base is somewhat limited, it takes a lot of communication and education for them to get committed to investing,” says Malay Ghatak, head of investments Emea at Citi.

“The spectacular issues in the market place around Ponzi schemes have left a number of investors shocked and disappointed. Clients are generally looking for a lot more transparency and they want to make sure there is a strong due diligence and robust process backing manager selection and ongoing performance management.”

“Quant strategies tend to work very well in a very directional market, but given we are in a choppy market, generally on the upside, they will have difficult time getting the right strategy in place,” says Cesar Perez, chief investment strategist for Emea at JP Morgan Private Bank. “We think quant strategies will underperform in 2011,” he says, explaining wealthy individuals have little exposure to them.

In quantitative strategies, it is important to employ a multi-factor approach to get to the final securities selection, explains Pierre Moulin, head of financial engineering at BNP Paribas Investment Partners. “We believe in a multi-factor approach because there is never one single driver of market returns. Markets are extremely complex, there is a lot of information that needs to be absorbed.”

That’s why it is important to use different types of non-correlated sources of information and diversify the time horizon of the specific information to explain the market. When correlations are high, like during the recent crisis when fear was the single driver of all asset class behaviour, the environment becomes very challenging for quantitative strategies, which exploit differences in valuation and fundamentals among securities within a given asset class. But the current climate offers better opportunity in the quant space, says Mr Moulin.

Quant strategies in general are non directional, with performance uncorrelated to the rest of the market. “My advice would be to allocate the risk budget to different types of quant approaches, because that is the best way investors can enhance their risk- return profiles,” he says.

One of the reasons these strategies may be difficult to explain to non professional investors is that often people only look at returns, not risks, whereas both should be taken into account, and one of the strengths of quant investing is strong risk control, says Jean-François Schmitt, CEO (UK) Quant, HSBC Global Asset Management. Very risky stocks can generate very strong expected returns, but they may be very volatile and maybe not very liquid.

“I am a strong believer that a systematic approach with risk control delivers on average better value added for the clients,” he says.